RSNA 2007 

Abstract Archives of the RSNA, 2007


SSG18-08

Computer-aided Detection (CAD) in Lung Cancer Screening at Chest MDCT: ROC Analysis of CAD versus Radiologist Performance

Scientific Papers

Presented on November 27, 2007
Presented as part of SSG18: Chest (Lung Nodules, CAD)

Participants

Francesco Fraioli MD, Presenter: Nothing to Disclose
Linda Bertoletti MD, Abstract Co-Author: Nothing to Disclose
Maria Luisa Mennini, Abstract Co-Author: Nothing to Disclose
Goffredo Serra, Abstract Co-Author: Nothing to Disclose
Carlo Catalano MD, Abstract Co-Author: Nothing to Disclose
Roberto Passariello MD, Abstract Co-Author: Nothing to Disclose

PURPOSE

Our aim was to compare the performance of radiologists against a CAD algorithm for pulmonary nodule detection on MDCT.

METHOD AND MATERIALS

Three radiologists independently analysed 250 scans and assigned to each nodule a confidence score (1-3). CAD was applied to all scans, successive readers re-evaluated all findings of the CAD, assigning, in consensus, a confidence score (1-3). The reference standard was established by other two experienced chest radiologists. Results were used to generate an FROC analysis.

RESULTS

The reference standard showed 139 nodules. Sensitivity for readers I-II-III was: 50, 62, and 38%. A double and triple reading resulted in an increase in sensitivity up to 74%. With CAD, sensitivity was increased to 93, 96, and 94% for readers I, II, and III. The area under the ROC curve (Az) was 0.55, 0.71, 0.44, and 0.82 for readers I, II, III, and the CAD. Differences between all readers and the CAD were significant (P 6 mm Az was 0.74, 0.90, 0,66 and 0.90 for readers I, II, III, and the CAD. P was significant between reader I and the CAD and between reader III and the CAD, it was not significant (P=0.9) between reader II and the CAD. For nodules < 6 mm, Az was 0.20, 0.21, and 0.19 for readers I-III and 0.65 for the CAD. Differences between all readers and the CAD were significant (P <0.05).

CONCLUSION

CAD can aid in daily radiological routine detecting a conspicuous number of nodules unseen by radiologists.

CLINICAL RELEVANCE/APPLICATION

With the recent introduction of the new class of 64-row CT scanners, CAD systems will have an important role for radiologists to efficiently review and interpret the vast amount of data offered.

Cite This Abstract

Fraioli, F, Bertoletti, L, Mennini, M, Serra, G, Catalano, C, Passariello, R, Computer-aided Detection (CAD) in Lung Cancer Screening at Chest MDCT: ROC Analysis of CAD versus Radiologist Performance.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5010449.html